Lakehouse RT vs. StarRocks

Summary

• Lakehouse Real-Time (Lakehouse//RT) is a Beta SQL warehouse type on Databricks serverless compute for low-latency, high-concurrency read queries on Unity Catalog managed Delta Lake and Apache Iceberg tables.
• StarRocks is an open-source (Apache 2.0) massively parallel processing (MPP) database for real-time analytics, with a vectorized engine, a cost-based optimizer and materialized views.
• Lakehouse//RT reads tables that Unity Catalog already governs; StarRocks stores data in its own tables or queries lake formats such as Iceberg and Delta Lake through external catalogs.
• StarRocks handles loading, updates and deletes itself; Lakehouse//RT runs ANSI read-only queries and leaves writes to other Databricks compute.
• Lakehouse//RT is managed by Databricks and enabled per workspace during the Beta; StarRocks is a cluster you deploy and operate, manually or on Kubernetes.

Lakehouse RT vs. StarRocks

Both serve low-latency analytics on fresh data, from different starting points. Lakehouse//RT is a Beta warehouse type inside Databricks that queries Delta Lake and Iceberg tables in place under Unity Catalog. StarRocks is an open-source analytical database that you deploy, operate and secure as its own system. If your data and governance already live in Unity Catalog, Lakehouse//RT serves it without a second store; if you want a standalone engine with its own storage and write path, StarRocks is built that way.

What is Lakehouse//RT?

Lakehouse Real-Time (Lakehouse//RT) is "a serverless compute built for low-latency, high-concurrency use cases", such as serving analytical data to custom applications, running operational analytics or powering BI dashboards. It's a SQL warehouse type: after a workspace admin enables the Beta under Previews, you create one under Compute > SQL Warehouses and select Real-Time.
It queries Unity Catalog managed Delta Lake and Iceberg tables, materialized views, streaming tables and metric views. Queries are ANSI read (SELECT) statements sent through the Statement Execution API or Databricks drivers such as JDBC and the SQL Connector for Python. The docs say its performance characteristics and supported features will change before general availability.

What is StarRocks?

StarRocks describes itself as a massively parallel processing (MPP) database designed to make real-time analytics easy for enterprises. Its features include a fully vectorized engine, a cost-based optimizer and intelligent materialized views, and it's compatible with MySQL protocols and standard SQL.
A cluster has frontend nodes (FEs) for metadata and query planning, plus backends. In shared-nothing mode, backend nodes store data locally. In shared-data mode, compute nodes keep only a cache, and data sits on object storage or HDFS. External catalogs for Hive, Iceberg, Hudi, Delta Lake and JDBC sources let StarRocks query external data without loading it. Primary Key tables support real-time updates and deletes, and Routine Load streams messages from Kafka topics. StarRocks manages privileges with role-based and identity-based access control.

How do Lakehouse//RT and StarRocks compare?

Dimension	Databricks Lakehouse//RT	StarRocks
What it is	Real-Time SQL warehouse type on Databricks serverless compute (Beta)	Open-source MPP database (Apache 2.0)
Where data lives	Unity Catalog managed Delta Lake and Iceberg tables in cloud storage	StarRocks tables on local disks (shared-nothing) or object storage and HDFS (shared-data)
Lake tables	Reads managed Delta Lake and Iceberg tables directly; external tables aren't supported yet	External catalogs for Hive, Iceberg, Hudi, Delta Lake and JDBC sources
SQL and clients	ANSI read-only SQL via the Statement Execution API and Databricks drivers	Standard SQL over MySQL protocols
Writes	None; other Databricks compute writes the tables	Loading methods, plus Primary Key tables for updates and deletes
Streaming freshness	Reads streaming tables and materialized views that Lakeflow Pipelines keep current	Routine Load from Kafka, plus Kafka and Flink connectors
Access control	Unity Catalog; no ABAC, row-level security or column masking in the Beta	Built-in role-based and identity-based access control
Deployment	Managed by Databricks, enabled per workspace in supported regions	Self-deployed, manually or on Kubernetes with the operator or Helm

How does Databricks approach real-time analytics?

Databricks launched Lakehouse//RT at Data + AI Summit in June 2026 to run real-time analytics directly on governed Delta Lake and Iceberg tables, without copying data into a separate serving system. Lakeflow Pipelines ingest from sources such as Apache Kafka and cloud storage into streaming tables and materialized views, and Lakehouse//RT serves reads on them.
The docs recommend validating queries on a serverless SQL warehouse first, using managed tables with predictive optimization and liquid clustering, keeping queries selective, and pre-aggregating complex joins with materialized views. When a query uses an unsupported feature, Lakehouse//RT returns an error naming that feature, and you run the query on a serverless SQL warehouse instead.

FAQs

Does StarRocks need its own copy of lake data?

Not always. StarRocks can load data into its own tables, and its docs say external catalogs query Hive, Iceberg, Hudi and Delta Lake data with zero data loading or migration. Lakehouse//RT reads Unity Catalog managed tables in place; external tables in Unity Catalog aren't supported yet.

What SQL does each one use?

StarRocks is compatible with MySQL protocols and supports standard SQL, so MySQL clients can connect. Lakehouse//RT runs ANSI-compliant read queries in ANSI mode only, and it doesn't support AI functions, Python UDFs, or the GEOGRAPHY and GEOMETRY types.

How does StarRocks handle real-time updates?

Through Primary Key tables. When a new row has the same primary key as an existing row, the new row replaces it. The docs describe syncing changes from transaction processing systems with tools such as CDC Connectors for Apache Flink.

Can Lakehouse//RT and Databricks SQL use the same tables?

Yes. Both query Unity Catalog tables, and when a query needs a feature Lakehouse//RT doesn't support, the docs direct you to run it on a serverless SQL warehouse instead.